Science Literature

Robert Deyes

Real life compounds and real life experiments were the themes of a Nature Biotechnology article about a biologically-based machine that could play Tic Tac Toe (In the UK, the game of Naughts and Crosses) against human players. A Nature review of the paper told of a machine that could make 'decisions' that would not only allow it to win but win every time (Ref 1). Tic Tac Toe is of course a very simple game to play the idea being that the first to generate a row of either three crosses or 'zeros' on a nine-square grid wins the game. This new version of the game on the other hand was different and involved a rather contrived set of operational rules. Designed by Milan Stojanovic and Darko Stefanovic the machine was made up of nine individual square holes or wells, rather like a waffle, each containing a special concoction of biological enzymes. Play would begin when the human player added magnesium to each well of the waffle, triggering off a series of reactions in the process (Ref 1,2). The machine would then 'make' its first move, always from the center square of the nine-square waffle visible through special enzymatic reactions that caused the well to change colour. The player would be forced to then make their first move in the top left most square and, in the process, add a special piece of DNA to all of the wells (Ref 1,2). Thus every well was 'informed' of the player's first move. And so it went on, each time the human player being obligated to add special pieces of DNA to each of the remaining wells. The most that the human player could ever hope for was a draw (Ref 2). Nevertheless, it seemed from Stojanovic's and Stefanovic's design plan that complex biological 'machines' could be designed to make decisions based on inputs fed into it by human agents. Artificial Life had apparently made a move out of the computer-based simulation.

In another sense, Stojanovic's and Stefanovic's game was altogether very disappointing for those keen to demonstrate a simple route for generating complex feedback networks. After all, this biological machine had no powers to gain from experience and improve its strategy. Moreover, unlike the human players, Stojanovic's and Stefanovic's game did not display any ability to self-organize thought processes into winning strategies or chemicals into closed sets of catalytic reactions. It simply did what it had been instructed to do and all this within a very constraining set of conditions. The inventory list of components that Stojanovic and Stefanovic had used, for example, had been carefully selected for what was conceptually a very simple game. Each hole of the nine-welled waffle contained carefully designed DNA enzymes with YES, AND and NOT gates rather like the electrical circuit boards used in high school physics classes. The conditions within each hole of the nine-well waffle were carefully tweaked and fine-tuned to ensure that the reactions would work. The authors themselves reported how their initial attempts at playing the game resulted in, "further empirical improvements" to ensure that the machine worked (Ref 2). Indeed this biological machine never lost because it implemented the perfect strategy. In all fairness, Stojanovic and Stefanovic never designed their machine to learn from experience. Nevertheless they were successful at demonstrating just how difficult it is to design a biological machine that could carry out even the simplest of tasks - hardly what one would hope for those arguing for the natural emergence of complex biological systems.

The same sort of bleak findings have been reported in other studies designed to harness the practical benefits of bacteria. Drew Endy, a biologist at MIT, is a pioneer in making novel bacterial strains that might one day become bioproducers of disease targeting drugs (Ref 3). Yet for Endy and other biologists, the results of genetic engineering have been disappointing and frustrating to say the least primarily because of the time needed to engineer bacteria to do even the simplest of tasks (Ref 3). Endy has taken on the challenge of designing a library of genetic parts, or 'biobricks', that can be inserted into bacteria to achieve certain basic functions. Endy's work might one day soon provide an accessory shop of sorts where researchers can customize their bacteria rather like automobile aficionados might customize their car. Yet the problem with customizing bacteria in this way is that, over relatively short periods of time, the genetically engineered parts mutate and become non-functional. As Ron Weissman of Princeton reported,

"Replication is far from perfect. We've built circuits and seen them mutate in half the cells within five hours. The larger the circuit is, the faster it tends to mutate" (Ref 3).

As science writer Wayt Gibbs reminds us, scientists are coming to terms with the relative difficulty of engineering even the shortest stretches of DNA to achieve even the most basic of functions (Ref 3). Many of the products of these short pieces of DNA have proved to be toxic to the cells requiring careful tweaking and trimming before being successfully introduced into bacteria. In other words, biological entities such as bacteria do not lend themselves easily to even the smallest of changes in their molecular composition. They require precision-directed design if such viable changes in their molecular constitution are to be made.

References

1. The Nature Review of the paper written by Stojanovic and Stefanovic was published by Helen Pearson in Nature on the 21st of August, 2003 and can be found at http://www.nature.com/nsu/030818/030818-9.html

2. Milan N Stojanovic and Darko Stefanovic (2003), A deoxyribozyme-based molecular automaton, Nature Biotechnology, Volume 21 Number 9 pp 1069 - 1074.

3. W. Wayt Gibbs (2004) Synthetic Life, Scientific American, Volume 290 (5) pp 74-81.